结合双通道WGAN-GP的多角度人脸表情识别算法研究  

Multi-Angle Facial Expression Recognition Algorithm Combined with Dual-Channel WGAN-GP

作　　者：邓源 施一萍[1] 刘婕 江悦莹 朱亚梅 刘瑾[1] Deng Yuan;Shi Yiping;Liu Jie;Jiang Yueying;Zhu Yamei;Liu Jin(School of Electronic and Electrical Engineering,Shanghai University of Engineering Science,Shanghai 201620,China)

机构地区：[1]上海工程技术大学电子电气工程学院,上海201620

出　　处：《激光与光电子学进展》2022年第18期127-137,共11页Laser & Optoelectronics Progress

基　　金：国家自然科学基金(61701296);上海工程技术大学学科建设项目(20KY0218)。

摘　　要：针对传统算法对多角度人脸表情识别效果不佳、偏转角下生成的人脸正面化图像质量低等问题,提出了一种结合双通道WGAN-GP的多角度人脸表情识别算法。传统模型仅利用侧脸特征对多角度人脸进行表情识别,特征差异小导致识别精度低。因此,引入生成对抗网络对人脸进行转正,消除姿态角的影响。为了使模型稳定训练的同时提升人脸生成质量,以WGAN-GP作为基础网络,并将其改进为双通道结构,融合五官特征及人脸全局特征来进行正面化生成。最后,构建轻量级网络MobileNetV3对生成出的正面人脸表情图像进行识别,保证分类精度并且大幅减小参数运算量。实验结果表明,所提算法在任意角度下,都能较好地复原出正面化人脸表情图像,提高了多角度人脸表情的识别率。A multiangle facial expression recognition algorithm combined with dualchannel WGANGP is suggested to address the concerns of poor performance of standard algorithms for multiangle facial expression identification and bad quality of frontal face pictures generated under deflection angles.Traditional models only use profile features to recognize the multiangle facial expression,which leads to low recognition accuracy due to small differences in characteristics.As a result,the generative adversarial network is used to frontalize the face first,removing the impact of the pose angle.To stabilize the training of the model and improve the quality of face generation,WGANGP is used as the baseline and improved into a dualchannel structure,which fuses the facial features and the global features of the face for frontalization.Finally,the lightweight network MobileNetV3 is built to detect the produced frontal facial expression photos,ensuring classification accuracy while drastically reducing parameter calculation.The experimental results demonstrate that the proposed method can well generate the frontal facial expression images at any angle and enhance the recognition rate of multiangle facial expressions.

关 键 词：图像处理 生成对抗网络 卷积神经网络 多角度人脸表情 人脸正面化 双通道 

分 类 号：TP391.[自动化与计算机技术—计算机应用技术]